CS 505 Computational Image processing

Fall 2006, Tues 4:10-7:05pm Room: TBD ?????

This course will be mixture of the underlying principles of image processing with a lot of the issues in making it practical. It will examine computational and theoretical aspects of image processing including representations, image formation/sensing, sampling theory, reconstruction/interpolation, warping and geometric transforms, filtering, wavelets and FFTs as well as higher level image processing concepts including segmentation, noise removal, and feature/target detection and tracking. While focused on images for examples, most of the computational techniques apply across many areas of signal processing. At least 1 assignment will be doing real-time image processing in C/C++ and one will be an spectral analysis in Matlab.

Overall I expect course will have 3-4 software assignments (maybe teams, maybe notm depends on the size of the class). No regular homeworks but maybe 1-2 quizes. Largely lecture but some times there will be discussion oriented classes as well. All students will give 1-2 presentations (depending on class size).

Using 2 texts :

• Image Processing in C by D. Phillips
• Feature Extraction & Image Processing Mark S. Nixon and Alberto S. Aguado, http://www.ecs.soton.ac.uk/~msn/book/
• And a free/oneline one:

  Hypermedia Image Processing Reference by R.B. Fisher, S. Perkins, A. Walker, E. Wolfart. John Wiley and Sons, 1996

Assignment #1, due Sept 21 7am (yes early morning thursday).

This assignment is 100 points 5% of your overall grade, so please take it seriously.
In http://www.vast.uccs.edu/~tboult/CS505/IMGS there 7 images (one pretty large) to which you are to use FFT analysis to try to remove the distortions/noise. For each one you are to do an openCV, Matlab or other program to "correct it". It is not sufficient to turn in the images, I want the steps you undertake, the code you produce and why that is the right solution (as comments in code is fine). If you make assumptions or discoveries along the way, make them clear in comments in the code both what they are and how you reached them.
For 25 bonus points, do at least 2 images, one of which should be the large image, in both openCV and matlab and discuss the timing differences and the ease of coding.

Assignment #2 (will be using matlab, doing restoration).. to be assigned around Nov 1)

Final project.

Your project is to do some applied computational image processing, building on something from the course. The subject of the project is up to you, but when thinking about the scale, keep in mind the expected effort is on the order of 50-60 hrs for programming and writing the report. (This is 40% of your course grade.)
The platform is up to you. The language is up to you. But the project has a number of constraints:

• Your project cannot be just the a reimplementation of something in a common image processing library. It may provide the same functionality, but must do it in a novel manner. Or it can provide a novel feature. Or it can moderately improve the performance of an existing implementation. It must add value, either in novelty or performance.
• You may use libraries, but must write at least most of the the "significant" functions for your project. The description of the functionality you will implement yourself must be in your project outline, and refined in your interim report.
• Your outline should describe the problem to be solved, the main image processing steps you see, how you will measure your progress, and the platform/libraries to be used.
• Your "testing" must involve at least a dozen images/examples to validate your project.